TECH NEWS – OpenAI’s new technology will benefit both Intel and AMD. For us, however, it will just mean higher prices again.

OpenAI hasn’t opened the “floodgates” for its latest GPT-6 Astra AI model yet, but early impressions on social media suggest the model creates many agents. This further reinforces the overarching theme of agent coordination and the associated demand for more CPUs. GPT-6 Astra is available to select customers and represents a significant shift in how large language models (LLMs) typically operate. OpenAI emphasizes that with Astra, users will never have to click a mouse or type on a keyboard again. Rather than requiring developers to provide separate APIs for every application an AI agent needs to use, GPT-6 Astra can navigate any software just as humans do. It creates agents that operate in browsers, spreadsheets, websites, and desktop applications. These agents can produce finished documents and presentations and execute multi-step workflows instead of simply telling users how to perform them.

GPT-6 Astra’s core reasoning engine uses a native, multi-agent structure to solve complex problems. When faced with a complex task, the main coordinating agent can formulate a theory and deploy sub-agents to test variations and validate results in parallel. This native task delegation makes the model extremely resilient to “death loops” (when it gets stuck in repetitive error cycles), enabling it to debug its own code and adjust its strategy independently. While the GPT-6 Astra core runs in the cloud, the model’s role as a native computing operator creates a computationally intensive loop directly on the local device. This offloads the workload to the host CPUs for three main reasons:

1. Given OpenAI’s emphasis on GPT-6’s ability to independently discover and link security vulnerabilities, companies will likely deploy the model in isolated local virtual environments, such as sandboxes and secure containers (Docker and MicroVM). Creating, maintaining, and shutting down these instances is an extremely CPU-intensive process. To feed Astra their own data, companies must run a test framework or coordination code on local machines, which also places a heavy CPU load on the system. When Astra launches subagents to test variations in parallel or perform local debugging of a software development script, the local CPU runs these test suites. If Astra runs a unit test suite, compiles code, or quickly refreshes browser processes to verify a workflow, the local CPU bears the brunt of the execution load.

All of this means that the demand for CPUs will skyrocket, which could bode well for Intel and AMD. GPT-6 Astra partially hides its thought process, making it difficult to uncover. This makes it (at least in theory) even harder for Chinese open-source AI models to catch up.

Source: WCCFTech